PyTorch CTO frames vLLM and SGLang as a cheaper way to buy useful AI work
PyTorch · x · 2026-07-23
At AMD's Advancing AI event, PyTorch Foundation CTO Matt White will outline a practical answer to a familiar problem: how to get more useful work, not just more tokens, from every AI dollar.
His talk focuses on open-source inference economics and the stack around vLLM and SGLang. The suggested operating model is to decompose workflows, use right-sized models, route and cache intelligently, escalate only when needed, and measure cost per successfully completed task rather than raw token volume.
Related event: PyTorch CTO to Discuss Open-Source AI Inference Economics(2 posts)→
More from Infra
- Nvidia Is Now Core to Every Major Robotaxi Stack at Commercial Scale — pdamodaran · 2026-09-11
- 12 KV Cache Reduction Techniques Every AI Engineer Should Understand, Explained — blaizedsouza · 2026-09-11
- The shadow GPU capacity market is formalizing, with Meta selling excess compute to outside buyers — DavidLinthicum · 2026-09-11
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11